Encrypted login | home

Program Information

Predicting Recurrence in Triple Negative Breast Cancer Patients From Clinical Parameters Using Different Classifiers

X Chen

X Chen1,2*, Z Zhou1 , K Thomas1 , M Folkert1 , N Kim1 , A Rahimi1 , J Wang1 , (1) UT Southwestern Medical Center, Dallas, TX, (2) Xi'an Jiaotong University, Xi'an, China


SU-H3-GePD-J(B)-1 (Sunday, July 30, 2017) 4:00 PM - 4:30 PM Room: Joint Imaging-Therapy ePoster Lounge - B

Purpose: To predict recurrence in triple negative breast cancer (TNBC) in stage II and III from clinical parameters by using different classifiers.

Methods: For this study, we used clinical information from 114 patients from our institution with TNBC with stage II-III. Patients were analyzed in three different groupings, all 114 patients, the subset of 80 patients who received neoadjuvant chemotherapy and the subset of 34 patients who did not receive neoadjuvant chemotherapy. A total of 35 clinical parameters were used as the candidate features for feature selection and predictive model construction. These parameters include 9 demographic parameters, 1 medication parameter, 6 tumor characteristics, 5 lab results, 4 staging parameters, 2 chemotherapy parameters, 1 radiation therapy and 7 surgery parameters. Sequential forward search (SFS), multi-objective optimization (MO), support vector machine (SVM) and logistic regression (LR) were used for feature selection and model training. MO classifier considers both sensitivity and specificity as the objective function simultaneously during feature selection and model training.

Results: The highest prediction accuracy was achieved when using MO+LR for feature selection and modeling. The prediction accuracy is over 80% for group 1 (all patients) and group 2 (patients with neoadjuvant chemotherapy). Among the selected features for these two groups, four clinical parameters were important for recurrence prediction: anticoagulant use, pathological stage grouping, tumor grade, and type of breast reconstruction.

Conclusion: We developed a MO+LR classifier for predicting recurrence in TNBC patients using key clinical parameters. This model may facilitate the stratification of high risk TNBC patients who are at highest risk for recurrence despite multimodality therapy.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by the Cancer Prevention and Research Institute of Texas (RP130109), the American Cancer Society (RSG-13-326-01-CCE), US National Institutes of Health (R01 EB020366) and the National Natural Science Foundation of China (61401349).

Contact Email: